The latest indyref polls have produced mixed results, with three confirming the No camp in the lead, and a fourth - with a smaller sample - suggesting Yes well ahead.

Survation, who were commissioned by Better Together to carry out a phone poll of 1000 Scots over the last 72 hours, reported 41% backing independence and 47% in favour of the union.A further 9% of voters said they were undecided and 3% refused to reveal voting intentions.

Stripping those categories out, the split was Yes 46% No 54%.

The poll also found 40% of voters believe they and their families would be financially worse off under independence compared to 27% who believe they would be better off.

The latest survey also points to a very high turnout, with 93% of voters saying they are certain to vote.

Meanwhile, an Opinium/Observer poll also showed a narrow gap in favour of No.

Excluding undecided voters, the poll of 1,055 adults between September 9 and 11 found 53% of those questioned plan to vote No while 47% intend to vote Yes.

The online surveyfound that many Scots see the Better Together campaign as being too negative (53%), lacklustre (54%), patronising (50%), and disorganised (49%).

The Yes campaign is regarded by some as too aggressive (43%) and unrealistic (46%), compared to 29% and 34% respectively for the Better Together campaign.

The study found that Alex Salmond and Nicola Sturgeon come out with the highest support on who the Scottish public think has done well during the campaign (51% for Mr Salmond and 48% for Ms Sturgeon), compared to 25% for Alistair Darling, 24% for Gordon Brown, 13% for David Cameron and 10% for Ed Miliband.

A Panelbase poll for the Sunday Times shows Yes at 49% and No on 51% once undecided voters are removed.

The fourth poll out today, by ICM for the Sunday Telegraph, puts Yes on 49% (up 11 points on the last ICM online poll) while No are on 42% (-5). Once the 9% of Don't Knows are removed, Yes are on 54%, up nine points, the highest vote for Yes ever recorded by ICM in one of its internet polls.

However, the survey sample was only 700, below the accepted minimum size of 1000, so the outcome could be unreliable.